DataFab / Utilities / Build your own
Governed utility · composed by you
Everyone ships an agent. We ship the studio.
A generic agent knows generic things. The utility that changes your business is the one nobody sells — because it encodes how your organisation specifically does the work. Any governed workflow can be composed in the Knowledge & Agentic Studio, grounded on the Fabric, and published as a reusable utility your other teams inherit.
Why build
The intelligence that matters is not for sale.
Six utilities ship ready. They cover the work that looks the same in every organisation in a sector. Everything else — the reconciliation nobody else has to do, the approval chain that exists because of a regulator you specifically answer to, the assessment your best person does in their head — is not a product anyone can sell you.
Encoded once, it becomes a governed, versioned object your organisation owns outright. That is the difference between renewing a licence and holding an asset.
The people who know the work
Domain discovery, planning from language and the visual builder mean the composers are the people who do the work — not a bench of specialists and not a change request.
Documents you already have
Specifications, policies, contracts and forms uploaded; the schemas extracted from them; the model derived rather than declared.
An afternoon, not a quarter
Describe the outcome, review the derived schema, test in a sandbox against sample files, publish scoped and gated.
A compounding library
Each utility your teams author is reusable supply the next team inherits — and every governed run enriches the Fabric beneath all of them.
How it is built
Compose, ground, govern, publish.
Agencies and widgets
Coordinated teams of agents, each with one job, plus the visual components the operator drives. Sequential, branching or hierarchical, with a human gate wherever judgement belongs.
On the Fabric graph
Every agent queries the resolved graph — entities, relationships and provenance — instead of raw, unresolved sources.
Sandbox, gates, audit
Isolated execution, tool authorisation scoped per use case, credentials injected at run and destroyed after, and a hash-chained trail with no delete capability.
A reusable utility
Draft to published, runtime placeholders mapped, access permissions scoped, and made available to everyone entitled to it.
What it inherits
Turning one on does not add a governance problem.
A utility you build is not a new application with its own data model, its own permissions and its own audit story to reconcile. It inherits all three from the layer beneath it.
The resolved entities
The same golden records, the same derived ontology, the same confidence and provenance on every edge.
The permission model
One consolidated model your organisation authors, enforced at traversal, with inherited entitlements as the floor.
The audit trail
Hash-chained, encrypted, auditor-role only, produced as the work happens rather than added afterwards.
The automation dial
Autonomy set per step, from deterministic through to fully agentic, so the control sits where the risk is.
Publishing
Built by your people. Owned by your firm.
Encode the expertise once, own it, and deploy it across every team, every client and every jurisdiction — running inside your boundary, on data that never moves.
Next step
Bring the workflow nobody sells a product for.
The one that spans the most systems, or the one your best person does in their head. That is the utility worth owning.